Diseño e implementación de un controlador neuro-difuso para un motor de corriente continua.

Abstract

The present work shows the design and implementation of a conventional P.I.D. controller, and a neuro-fuzzy P.I.D. controller, both of them designed for a direct current motor, the work begins by identifying and obtaining a mathematical model for the system, later a rst parameters tuning is made using the Ziegler-Nichols criteria and a optimization using the descending gradient technique. A neuro adaptative net ANFIS is used for the training and optimization of the weights for the fuzzy controller by using also the descending gradient technique. On both cases, for the optimization process a cost function was designed, by using the integral absolute error, the integral of the rate of change of the error, and a special function created for making the controller phisically doable, this special function can be marked as the most important contribution of this work because it generates a model that can be implemented not only in theory but in practice also. The results show in the theorical design phase and in the implementation phase with the measurements obtained that a neuro-fuzzy controller can generate responses beyond the limits of a conventional controller. By showing a diminishing of 70% in the rise time as a diminishing of 55% in the settling time with the generation of a very small over-shot in the system's temporal reaction.